The first scheme associated with packet transfer, in which a VC is set between adjacent node apparatuses and packets are transferred over the VC, uses a router incorporating an interface capable of setting an ATM (Asynchronous Transfer Mode) VC or the like. The router performs exchange processing for an incoming packet, determines the output destination of the packet, and outputs the packet over a VC corresponding to the determined output destination.
When, for example, an ATM interface is used, an ATM VC is set between adjacent routers, and capsulation processing defined by RFC 1483 (Internet Engineering Task Force (IETF) Request For Comments (RFC) 1483) is performed for a packet on the VC. The resultant packet is segmented into cells by using AAL5 (ATM Adaptation Layer Type 5), and the cells are transferred. The router reassembles the incoming packet, segmented into cells, into the original packet through the ATM interface, and performs exchange processing for the packet to determine an output destination. The router then segments the packet into cells through the ATM interface, and outputs the cells by using a VC corresponding to the determined output destination.
In the first scheme associated with packet transfer, the implementation of RSVP (Resource ReSerVation Protocol) (IETF RFC2205) makes it possible to perform quality control. According to RSVP, before a terminal which demands a quality guarantee for a transmission flow starts transmission, information for identifying the flow, e.g., a combination of terminals for performing transmission to all the routers on the route of the flow, a fourth-layer protocol type, and source and destination fourth-layer port numbers, and a table that describes a quality with respect to the flow are generated. With regard to a packet transmitted from the terminal, a router on the route of the flow identifies the packet in the flow by using the table, and performs exchange processing for the packet with a proper quality by using a quality guarantee mechanism.
As the second scheme associated with packet transfer, the scheme disclosed in Japanese Unexamined Patent Publication No. 10-56452 is available. According to this scheme, a node apparatus having a packet exchange function added to an ATM switch is used. In this scheme, a packet is segmented into cells by using AAL5 to be transferred over a VC as in the first scheme. However, a characteristic feature of this scheme is that each node apparatus directly performs exchange processing for the packet segmented into cells, and hence need not reassemble the cells into the packet. When a cell of the packet arrives at the node apparatus, it is checked whether the cell is the start cell of the packet. If this cell is the start cell of the packet, an output destination for the packet formed by the cell is determined from the packet header in the cell, and all the cells from the start cell to the final cell of the packet are consecutively sent out to the output destination.
The first problem associated with packet transfer is that a node apparatus in the second scheme associated with packet transfer has no function for quality guarantees. For this reason, even if the quality of a VC serving as a transmission route is guaranteed, since the node apparatus equally handles all traffics, e.g., high-priority traffics, low-priority traffics, and traffics requiring a real-time feature, in exchange processing, no quality can be guaranteed for high-priority traffics and traffics requiring a real-time feature.
The second problem associated with packet transfer is that in the second scheme associated with packet transfer, adjacent nodes are connected through one VC, traffics requiring different qualities must be transmitted over the same VC. For this reason, a VC must be set to guarantee the qualities of all traffics. As a consequence, a VC is set in accordance with a traffic demanding the highest quality. According to these schemes, therefore, a high quality is guaranteed for even a traffic requiring a low quality, wasting network resources.
The third problem associated with packet transfer is that the first scheme associated with packet transfer requires a large table to identify a flow, and hence is difficult to apply to a large-scale network. This is because pieces of information for uniquely identifying flows, e.g., combinations of terminals that perform communications through all the flows, fourth-layer protocol types, and source and destination fourth-layer port numbers, must be set in a table in units of flows. In the first scheme associated with packet transfer, since quality is guaranteed in units of flows, flexible quality guarantees, e.g., a quality guarantee for each terminal, a quality guarantee for each LAN (Local Area Network), and a quality guarantee for each virtual dedicated network, cannot be performed.
Recently, the IP traffic in a public network has sharply increased. For this reason, it has become urgent to provide high-speed, broadband IP services on an ATM network widely used in the public network. As schemes of processing packets on the ATM network, the following schemes are known.
In the first scheme of handling packets on the ATM network, a high-speed IP router incorporating an ATM line is set on the ATM network, and packets are exchanged by the IP router. More specifically, a packet that is input after it is segmented into cells is temporarily reassembled into the packet, and then exchange processing is performed. After the processing, the packet is segmented into cells again and output. According to the first scheme, a VC (Virtual Channel) is set between adjacent IP routers, and capsulation processing defined by RFC 1483 (Internet Engineering Task Force (IETF) Request For Comments (RFC) 1483) is performed. Thereafter, the packet is segmented into cells according to AAL5 (ATM Adaptation Layer Type 5) and transferred.
In the second scheme of handling packets on an ATM network, an IP layer processing function is added to an ATM switch itself to use the ATM switch as a node apparatus equivalent to an IP router. According to this scheme, a processing section equivalent to an IP router is added to an ATM switch, and the processing section performs IP layer switching for cells constituting a packet. A cell that has arrived at this ATM switch is identified as a cell of this packet or not. If this cell is a cell of the packet, the cell is transferred to the processing section. The processing section reassembles cells into a packet, i.e., restores packet pieces as cells into an original packet, and performs IP layer switching. A packet having undergone this processing is segmented into cells again. These cells are returned to the ATM switch section and output to the next node.
In the third scheme of handling packets on an ATM network, an IP layer processing function is added to the ATM network itself to allow the overall ATM network to implement an IP router function. As this scheme, first of all, SMDS (Switched Multi-megabit Data Services) is available (Japanese Unexamined Patent Publication No. 6-62038). SMDS is a connectionless traffic accommodation scheme developed by Bellcore in U.S.A. This scheme is effective in accommodating IP traffics in an ATM network. According to SMDS, a header and trailer are added to a packet at a node existing in a peripheral portion of an ATM network, and the packet is segmented into cells. The cells are then sent to the ATM network. A characteristic feature of SMDS is that switching is performed in each switch in the ATM network without reassembling cells into a packet, and hence apparatuses are simplified. In addition, since multiplex identifiers are added to the respective segmented cells, when cells constituting different packets are multiplexed on the same CV, no problem arises even if a cell of a packet is inserted in a cell group constituting another packet.
As another scheme associated with the third scheme of handling packets on an ATM network, the scheme disclosed in Japanese Unexamined Patent Publication No. 10-56459 is available. This scheme will be referred to as the fourth scheme. In the fourth scheme as well, switching is performed without reassembling cells into a packet. However, no multiplex identifiers are added to the cells of the packet, and the packet is segmented into cells by using AAL5. More specifically, a node apparatus has a first table holding a predetermined destination VPI/VCI (Virtual Path Identifier/Virtual Channel Identifier) and an output port number for each destination IP address, and a second table for holding a destination VPI/VCI and output port number for each source VPI/VCI as needed. When the start cell of an AAL5 frame associated with a given packet is received, a destination VPI/VCI and output port number are obtained by searching the first table by using the destination IP address in the cell as a key. After the VPI/VCI in the start cell is converted into the obtained destination VPI/VCI, the cell is output from the obtained output port number. At the same time, a matching source VPI/VCI is obtained by searching the second table by using the VPI/VCI in the start cell as a key, and the destination VPI/VCI and output port number obtained from the first table are held in the second table in correspondence with this source VPI/VCI. When a cell other than the start cell of the AAL5 frame associated with the packet is received, a destination VPI/VCI and output port number corresponding to a matching source VPI/VCI are obtained by searching the second table by using the VPI/VCI in the cell as a key. After the VPI/VCI in the cell is converted into the obtained destination VPI/VCI, the cell is output from the obtained output port number.
Both node apparatuses in the first and second schemes of handling packets on an ATM network process a packet after input cells are temporarily reassembled into the packet, and output the processed packet upon segmenting it into cells again. A delay in transfer increases due to the processing time required for packet reassembly and segmentation, and the apparatus is complicated.
In the third scheme of handling packets on an ATM network, since switching for a packet segmented into cells is performed without packet reassembly, a transfer delay can be decreased, and the apparatus can be simplified accordingly. In the third scheme, however, since a packet is not transferred after it is segmented into cells by AAL5, the use efficiency of a band is poor. This is because in the third scheme, a 2-byte header and 2-byte trailer including a multiplex identifier, CRC, and the like must be set in the payload of each cell.
In contrast to this, in the fourth scheme of handling packets on an ATM network, since a packet is transmitted after it is segmented into cells by AAL5, and switching is performed without reassembling the cells into the packet, the problem in the third embodiment does not arise. In the fourth scheme, since packets arriving from different VCs cannot be multiplexed on the same VC, many VC resources are required. This point will be described in detail below.
According to the fourth scheme of handling packets on an ATM network, as described above, by searching the first table with the destination IP address contained in the start cell of an AAL5 frame associated with a given packet, a destination VPI/VCI after conversion of the VPI/VCI in each cell of the packet and a destination output port are determined. For this reason, when a plurality of packets having the same destination IP address are simultaneously received, the VPI/VCIs of the respective cells are converted into the same destination VPI/VCI, and the cells are output from the same output port. If, therefore, the cells of the respective packets are output to the same output VC, cells of a given packet are inserted in the cell group constituting another packet. Since the VPI/VCIs in the cells of the respective packets are the same, the cells cannot be reassembled into the original packets at the output destination. For this reason, in practicing the fourth scheme, when different packets arriving from different VCs are to be output to the same path, different VCs must always be used, requiring a large amount of VC resources.